Predicting intention to follow online restaurant community advice: a trust-integrated technology acceptance model

Purpose – This research investigates consumer intention to follow online community advice. Applying the technology acceptance model (TAM) to the context of online restaurant communities, the study empirically examines the effects of perceived usefulness, perceived ease of use, attitude and trust on the intention to follow online advice. Design/methodology/approach – The data were collected from 360 members of online restaurant communities on Facebook and analyzed using structural equation modeling (SEM). Findings – The findings revealed that trust, perceived usefulness and attitude are key predictors of the intention to follow online restaurant community advice. Originality/value – Extant research on the influence of online reviews on consumer behavior in the restaurant industry has largely focused on the characteristics of the review, reviewers or readers. Moreover, other studies have investigated consumers’ motivations to write online restaurant reviews. This study, however, takes a different approach and examines what drives consumers to follow the advice from online restaurant communities.


Introduction
The advent of Web 2.0 and the ensuing growth of user-generated content (UGC) have transformed the way individuals search for and share information. The interactive nature of social media has given consumers opportunities to communicate with firms and with one another, creating UGC (Arora and Sanni, 2019). In computer-mediated environments, consumers share opinions, knowledge and personal experiences by posting reviews of products, services and brands on social networking sites (SNSs), third-party review sites or firms' own websites (Kannan and Li, 2017). Before purchasing, consumers evaluate products by checking online reviews as a trustworthy information source (Filieri et al., 2018).
Online reviews hold an undeniable importance in the hospitality industry (Liu and Park, 2015), and in the restaurant sector in particular (Dixit et al., 2019). Because of the intangible, experiential nature of hospitality services, it is challenging to evaluate their quality before consumption (Liu and Park, 2015;Ruiz-Mafe et al., 2020). This uncertainty felt by consumers in their decisions causes them to seek information (Yang et al., 2018).
The paper is structured as follows: the next section provides a review of the related literature from which the research hypotheses are developed. Then, the research methodology is presented. The findings of the study are reported next. The final section discusses the results and their implications and concludes with research limitations and future research.
2. Theoretical background and hypotheses development 2.1 Online reviews eWOM exists in online consumer opinion platforms, online communities and SNSs (Cheung and Lee, 2012;Litvin et al., 2008). eWOM in the form of online reviews include consumers' unbiased advice, opinions and comments (Hern andez-Ortega, 2020). Online review is defined as "any positive, neutral, or negative online review about a product or service created and published on a consumer review website by a potential, former, or actual customer." (Filieri, 2015(Filieri, , p. 1262. Reviewing the recent hospitality literature on online reviews, particularly on restaurants, we concluded that there are two streams of research. The first stream has examined the impact of online reviews on consumers' responses to the review (e.g. review credibility and helpfulness) or other behavioral responses (e.g. purchase intention) using three factors: source characteristics (reviewers), characteristics of the review (message) and characteristics of information receivers (review readers) (Zhang et al., 2021). The second research stream has focused on investigating consumers' motivations to write online restaurant reviews and drivers of consumers' eWOM intentions.
Concerning source characteristics in the first research stream, researchers have studied the following: the effect of reviewers' identity disclosure, reputation, expertise, age, gender and contribution badge on the helpfulness/usefulness of online reviews (Liu and Park, 2015;Srivastava and Kalro, 2019); the influence of source trustworthiness, expertise and homophily on consumers' purchase intentions (Filieri et al., 2018); the impact of source credibility on information diagnosticity (Filieri, 2015) and the moderating effects of reviewer's experience, cognitive effort in writing a review and online status on online review rating behavior (Li et al., 2020).
Regarding review characteristics, studies have investigated the following: the effect of textual features (i.e. attribute salience, review valence and content concreteness) on the trustworthiness of consumer reviews (Huang and Liang, 2021); the impact of positive valence intensity on consumers' post consumption responses, namely, attitude and repurchase intention (Hern andez-Ortega, 2020) and the influence of manifest/quantitative factors (i.e. review rating, word count and review length) and the latent/qualitative content factors (i.e. argument quality and message valence) on review helpfulness (Srivastava and Kalro, 2019).
Moreover, research on review characteristics has showed that eWOM metrics (e.g. volume, valence and composite valencevolume and variance) influenced the sales of products and services (Babi c Rosario et al., 2016;Yang et al., 2018;Ye et al., 2011;You et al., 2015). Similarly, the literature revealed that review volume and ratings positively influenced restaurant profitability  and online popularity (Zhang et al., 2010). Furthermore, some studies have analyzed the pictorial content of online reviews and its influence on purchase intention (Bigne et al., 2020) and perceived helpfulness of UGC (Bigne et al., 2021). Another group of studies have focused on the effect of positive and negative emotions embedded in online reviews on their perceived helpfulness (Chen and Farn, 2020;Ismagilova et al., 2020b).
Finally, from the information receivers' perspective, Zhang et al. (2021) examined how readers' construal levels influenced their perceived usefulness of online reviews. Further, Ruiz-Mafe et al. (2020) analyzed the intention of TripAdvisor users to follow its advice through the interrelationships between the cognitive cues of online reviews (i.e. perceived helpfulness and persuasiveness) and readers' emotions (i.e. pleasure and arousal). Results showed that review persuasiveness and helpfulness had positive effects on reader's pleasure and arousal, which determined the intention to follow TripAdvisor advice.
In the second stream, researchers in the restaurant industry have paid significant attention to studying consumers' motivations to write online reviews and the drivers of eWOM intentions (e.g. Dixit et al., 2019;Jalilvand et al., 2017;Yang, 2017). For instance, Dixit et al. (2019) reported that ego involvement, taking vengeance, perceived behavioral control and subjective norms predicted the intention to write online restaurant reviews. Furthermore, Jalilvand et al. (2017) identified four factors (i.e. food quality, personal interaction quality, physical environment quality and perceived value) that influenced eWOM intentions through relationship quality dimensions. Moreover, Yang (2017) emphasized that perceived usefulness and altruism had significant effects on restaurant customers' eWOM intention.
2.2 Technology acceptance model TAM proposed that a user's decision to adopt a technology is influenced by their intention. The intention is jointly determined by attitudes toward using the technology and perception of its usefulness. Two major beliefs about a new technology, namely, perceived usefulness (PU) and perceived ease of use (PEOU), jointly determine a user's attitude toward adopting it. Further, TAM hypothesizes that PEOU is an antecedent of PU .
In the hospitality sector, researchers have applied TAM to examine user acceptance and use of technology-related applications, including adoption of online reviews (Ayeh et al., 2016), intention to participate in online community (Agag and El-Masry, 2016), restaurant eWOM intentions (Yang, 2017) and restaurant visit intentions (Salehi-Esfahani and Kang, 2019).
Although the hospitality literature confirms the importance of PU and PEOU in predicting technology acceptance, the results differ on their relative influence on attitude and intention (Ukpabi and Karjaluoto, 2018;Wang and Jeong, 2018). Yang (2017) found that PU was the strongest determinant of intention to spread eWOM about restaurants, with PEOU having an insignificant effect on intentions. In addition, Salehi-Esfahani and Kang (2019) reported that PU influenced attitudes toward restaurant review websites more strongly than PEOU. In contrast, Ayeh et al. (2016) showed that PEOU had a stronger effect on attitudes, whereas PU had a stronger impact on intentions.
PU is one of the two beliefs of TAM that determine user attitude . Wang and Jeong (2018) argued that enhancing the efficiency and effectiveness of Airbnb led to positive attitudes toward it. Likewise, PU was a major predictor of customers' attitudes toward social media restaurant reviews (Popy and Bappy, 2020). Moreover, Salehi-Esfahani and Kang (2019) found that PU of restaurant review websites positively influenced attitudes toward them. Similarly, we expect that: H1. PU has a significant positive effect on attitude toward online advice.
PU has a direct effect on intention . The authors argued that employees developed intentions toward using job-related technologies that were instrumental in achieving desired outcomes regardless of their attitudes. An extensive body of the literature has supported this relationship. Particularly, Yang (2017) found that consumers were encouraged to engage in eWOM on restaurant review websites if they perceived them as useful. Moreover, Hajli (2018) found that social media enhanced information usefulness, leading to social WOM adoption. Therefore, we propose that: H2. PU has a significant positive effect on intention to follow online advice.
The literature has emphasized PEOU's role as a key determinant of PU. For example, Liu et al. (2017) and Salehi-Esfahani and Kang (2019) found that PEOU of restaurant review websites influenced customers' usefulness perceptions. Thus, we hypothesize that: H3. PEOU has a significant positive effect on PU of online restaurant community. Davis et al. (1989) found that PEOU was less influential than PU in affecting attitude. Previous studies came to the same conclusion (e.g. Ayeh et al., 2013;Nedra et al., 2019). In contrast, Popy and Bappy (2020) found that PEOU was stronger than PU in predicting attitudes toward social media restaurant reviews. We propose that: H4. PEOU has a significant positive effect on attitude toward online advice.
The path between PEOU and intention is not one of the proposed paths of TAM. However, King and He (2006) found that the relationship between PEOU and intention was significant in Internet applications, highlighting its importance in determining online intentions. Ayeh et al. (2013) confirmed this finding as PEOU had a significant direct effect on intention to use consumer-generated media (CGM). In contrast, Yang (2017) found that PEOU of restaurant review websites did not influence eWOM intentions. Accordingly, this relationship requires further investigation.
H5. PEOU has a significant positive effect on intention to follow online advice.
Attitude toward using a technology was included in the original TAM; however, Davis et al. (1989) reported that attitude was a partial mediator of beliefs on intention. This led to its removal as they found that the three-construct model was more parsimonious and equally powerful. Nevertheless, Rondan-Cataluña et al. (2015) argued that attitude becomes more powerful when using TAM to study voluntary intentions. Moreover, Ingham et al. (2015) argued that attitude played a more vital role in explaining consumer behavior than it did in work-related technology acceptance. Previous studies found that attitude explained the intention to use Instagram (Nedra et al., 2019), Airbnb (Wang and Jeong, 2018), restaurant review websites (Salehi-Esfahani and Kang, 2019) and social media restaurant reviews (Popy and Bappy, 2020). Thus, we hypothesize that: H6. Attitude has a significant positive effect on intention to follow online advice.

Trust
Individuals with shared interests come together in online communities to exchange knowledge in a regular and organized way over the Internet (Ridings et al., 2002). They are regarded as valuable and credible repositories of information for consumers and as influential marketing and promotion channels for firms (Li and Chang, 2016). The shared interest encourages members to reveal information about their preferences that traditional marketing fails to capture (de Valck et al., 2009). The importance of trust in online communities stems from certain factors characterizing them (Ridings et al., 2002). Members cannot communicate face to face, making it difficult to expect how others might behave. Thus, dependence on trust is more crucial online as it can mitigate uncertainty (Li and Chang, 2016). Trust has a profound effect on the success of online communities as it is the driving force for more participation and engagement (Hajli, 2018).
Trust is mostly measured as a set of beliefs about the trust object (Ingham et al., 2015). Oliveira et al. (2017) found that ability, benevolence and integrity explained a substantial sum of variance in trust. They are the most commonly used beliefs in trust research (Mayer et al., 1995). Thus, trust here is examined as a single-dimensional construct reflecting beliefs in the competence, benevolence and integrity of an online community.
The hospitality literature has confirmed the positive effect of trust on relationship commitment between hospitality SNSs and members (Li and Chang, 2016), restaurant customers' attitudes (Popy and Bappy, 2020), eWOM credibility and usefulness (Ismagilova et al., 2020a), and restaurant visit and recommendation intentions (Anaya-S anchez et al., 2019).
Particularly, research on trust in online communities has investigated drivers of website trust (e.g. Anaya-S anchez et al., 2019;Filieri et al., 2015). For instance, Filieri et al. (2015) revealed that information quality, Website quality and customer satisfaction significantly influenced trust toward CGM websites. Moreover, scholars have examined the consequences of trust as Popy and Bappy (2020) showed that trust in social media restaurant reviews positively influenced attitude toward them. In addition, studying tourists' trust in review websites while traveling, Anaya-S anchez et al. (2019) found that website trust explained the intention to visit and to recommend a restaurant online and offline.
Focusing on the trust-attitude relationship, Casal o et al. (2011) found that trust influenced attitudes toward online travel community advice. Similarly, Ayeh et al. (2013) combined TAM with trustworthiness to investigate travelers' intention to use CGM. Results showed that trustworthiness strongly influenced attitude. Furthermore, Agag and El-Masry (2016) reported a significant path from trust in online travel community to attitude toward it. Consequently, we hypothesize that: H7. Trust has a significant positive effect on attitude toward online advice. McKnight et al. (2002a) confirmed that trusting beliefs in e-vendors predicted consumers' intention to follow their advice. However, there is a scarcity of research on the relationship between trust and eWOM adoption (Ismagilova et al., 2020a). Among the few studies that examined this relationship, Casal o et al. (2011) revealed that trust positively influenced the intention to follow online travel community advice. Likewise, Jalilvand et al. (2017) revealed that trust toward restaurants positively influenced customers' WOM intentions. In contrast, Ayeh et al. (2016) showed an insignificant relationship between trust and online reviews adoption intentions. We propose that: H8. Trust has a significant positive effect on intention to follow online advice.
Based on the developed hypotheses, Figure 1 demonstrates the study's conceptual model.

Sampling and data collection
Because of the absence of dedicated restaurant review websites (e.g. Yelp) in Egypt, consumers depend mainly on reviews on SNSs. SNSs are the second most used channel for brand research, whereas recommendations and comments on social media are the second most used source for new brand discovery in Egypt (Hootsuite and We Are Social, 2021a). Furthermore, Egypt has witnessed increasing rates of social media adoption. Compared with a global rate of 54%, social media penetration in Egypt stood at 47% in 2021 (Hootsuite and We Are Social, 2021a). As of 2021, there are 45 million Facebook users (Hootsuite and We Are Social, 2021b), making it the most popular SNS in Egypt (Alexa, 2021).
SNSs can provide venues for the creation of online communities (Le, 2018). Accordingly, this research focuses on third-party online communities on Facebook or Facebook pages for restaurant reviews. Food enthusiasts create pages that are dedicated to sharing their restaurant experiences. Any Facebook user can be a member/fan of these communities by simply "liking" the page. Hospitality studies have devoted significant attention to online reviews posted on online platforms such as TripAdvisor and Yelp to examine their influence on consumers' decisions (e.g. Huang and Liang, 2021;Ruiz-Mafe et al., 2020;Srivastava and Kalro, 2019). However, few studies have examined online reviews posted on SNSs like Facebook, despite their importance as an eWOM source (Ladhari and Michaud, 2015).
Convenience sampling was used to collect data from three of the most popular restaurant review pages in Egypt, namely, Egyptian Foodies, Engezni and elmenus. The reviews come from two sources. The page creators write reviews based on their experiences. In addition, they encourage members to send their restaurant reviews and recommendations to be featured on the platform and shared with their 2.5 million fans to help improve dining decisions. The reviews combine textual and visual information (photos/videos).
As the scales used to operationalize the constructs are adopted from multiple sources, we performed a pretest to refine the scale items (Hair et al., 2010). Furthermore, we asked a few academics to review the questionnaire.
For the pretest, data were collected from undergraduate and postgraduate university students using self-administered questionnaires. They were chosen as 60% of Facebook users in Egypt are aged between 18 and 34 (Statista, 2021). Thus, this age group makes up roughly two-thirds of the Facebook users in Egypt. In addition, previous eWOM research often used a sample of university students to collect data (e.g. Chen and Farn, 2020;Zhang et al., 2021).
Only students who are members of the selected pages were asked to complete the questionnaire and provide feedback on the items' wording. A total of 47 responses were collected, and the data were analyzed using SPSS version 25. Cronbach's alpha was computed to estimate reliability. Results showed that all constructs demonstrated acceptable internal consistency. Accordingly, all items were kept for further analysis.
The required sample size for structural equation modeling (SEM) depends on several factors including model complexity (the number of constructs and their items) and item communality (the square of standardized loadings) (Hair et al., 2010). They suggested a minimum sample size of 300 for models with seven constructs or less and item communalities below 0.45. Given that there are five constructs in our model, each represented by four or more items, we decided to collect more than 300 questionnaires to account for any incomplete responses.
For the main study, data were collected using a web-based survey developed using KwikSurveys. The questionnaire was designed in English. However, as the study is  (Brislin, 1970). An introduction was provided to explain the research purpose to participants. Participation was voluntary where links to both versions were posted on the three pages. A screening question was used to ensure that only the page members can answer the survey. The questionnaire was divided into two sections. The first included statements used to measure the research constructs. The second section contained questions about gender, age group, reading review frequency and membership length. Within two months, we received 540 online responses. The final sample consisted of 360 useable questionnaires, whereas 180 incomplete responses were discarded. This translated to a response rate of 66.67%.

Measurements
The constructs were measured using scales adapted to the context of online restaurant communities (see Appendix). PU was measured using a six-item scale from Davis (1989) and . All items were rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The five-point scale format is employed to minimize the respondents' level of frustration, resulting in a higher response rate and quality (Babakus and Mangold, 1992).

Results
The profile of the sample is as follows: females comprised 73.3% of the sample. Most participants (54%) were in the age group of 21 to less than 31 years and 25% were 31 years and above. In addition, 34.4% read reviews "every few days," whereas 22.8 and 21.1% read reviews "several times a day" and "once a day," respectively. Further, 61.1% of respondents had been members of their preferred community for more than a year.
The data were analyzed using SEM. A two-step approach was employed in this study (Anderson and Gerbing, 1988). The first step used confirmatory factor analysis (CFA) to evaluate the measurement model. The second step involved examining the structural model.

Measurement model
To test for normality, skewness and kurtosis were computed for all items. Following Kline's (2011) guidelines, values in Table 1 showed no departure from normality.
A CFA was performed for each construct using AMOS 22 based on the maximum likelihood estimation method. Following Schumacker and Lomax's (2010) suggestion, PU, PEOU and trust models were specified to obtain a better model fit. Based on the CFA results, eleven items were removed (Appendix). The remaining items were retained for further analysis.
The scales showed an acceptable fit to the data (Table 2) (Hair et al., 2010). All latent constructs exhibited acceptable levels of reliability and validity. Cronbach's alpha ranged from 0.74 to 0.86, whereas composite reliabilities (CR) were between 0.74 and 0.87, higher than the 0.70 threshold (Hair et al., 2010). Moreover, convergent validity was established as all items' loadings were statistically significant and above 0.50 (Hair et al., 2010). In addition, the average variance extracted (AVE) equaled 0.5 or higher, except for PU (0.42). However, AVE is a more conservative measure of convergent validity than CR (Fornell and Larcker, 1981). Based on CR values alone, we can conclude that a construct's convergent validity is adequate.
Finally, an evidence of discriminant validity was provided as the AVE values of any two constructs were higher than the squared correlation between them (Fornell and Larcker, 1981). Thus, each construct correlated more strongly with its own items than with any other construct in the model. Table 3 presents the squared correlation coefficients among the constructs and their AVEs.

Structural model
Overall model fit analysis indicated a satisfactory fit (Table 4). Once an acceptable fit is obtained, SEM is used to determine the model's explanatory power, size of path estimates and significance of hypothesized paths (Ayeh et al., 2013). The model explains intention to follow online community advice at a high level, accounting for 74% of its variance. Moreover, it explains 51% of the variance in attitude and 30% of the variance in perceived usefulness. Table 4 shows results of hypothesis testing by SEM.

Discussion and conclusion
This research proposed and empirically tested a conceptual framework to explain consumer intention to follow online restaurant community advice. Based on the literature, identifying the antecedents of following an online advice calls for a model that combines the following: (1) characteristics of the technology facilitating the knowledge exchange (TAM) and (2) factors related to the uncertainty in the online environment (trust). Therefore, we proposed a trust-integrated model based on the TAM to accomplish that objective. The results revealed several key insights. First, as hypothesized, perceived usefulness and perceived ease of use had significant effects on attitude toward following online advice. Although an easy-to-use online platform leads to favorable attitudes toward its advice, consumers place more importance on useful information. Similarly, Salehi-Esfahani and Kang (2019) found that perceived usefulness of restaurant review websites influenced users' attitudes more strongly than perceived ease of use. In contrast, Popy and Bappy (2020) showed that perceived ease of use was more critical in forming attitudes toward using social media reviews for choosing restaurants.
Our findings revealed that perceived usefulness was a direct antecedent of intention . Specifically, consumers are more likely to follow online restaurant community advice if they think it is useful. This result supports Liu et al. (2017) and Yang (2017), who showed that perceived usefulness determined restaurant review websites' adoption intentions and restaurants' eWOM intentions, respectively. Furthermore, perceived ease of use had its strongest influence on perceived usefulness, suggesting that it is a key determinant . This implies that the ease of obtaining valuable restaurant advice from online communities drives members' usefulness perceptions. This finding mirrors past restaurant research (Liu et al., 2017).
Contrary to prediction, perceived ease of use showed an insignificant negative effect on intention. This can be justified by the fact that 61% of the respondents had been members of online communities for more than a year. Hence, they were familiar with them, and usability did not influence their intentions. In fact, Ashraf et al. (2014) argued that perceived ease of use influenced intention when users did not have much experience with a technology but failed to do so when they became familiar with it. Similarly, Yang (2017) highlighted the insignificant effect of perceived ease of use on restaurant eWOM intentions. Conversely, Liu et al. (2017) supported its direct influence on consumers' intention to adopt restaurant review websites. Moreover, our findings confirmed that attitude predicts voluntary intentions (Rondan-Cataluña et al., 2015), which is consistent with prior restaurant research (Popy and Bappy, 2020;Salehi-Esfahani and Kang, 2019).
Finally, our findings indicated that trust in online community is a crucial factor that leads to positive attitude toward following its advice, reinforcing Popy and Bappy's (2020) results in the context of social media restaurant reviews. Further, results revealed that trust in online communities is a significant direct determinant of the intention to follow online restaurant community advice.
In this study, trust influenced intention more strongly than TAM constructs and was the second strongest determinant of attitude in the context of online restaurant communities. Comparing our result with previous hospitality research revealed contrasting findings. Reviewing UGC adoption in travel planning studies, Ukpabi and Karjaluoto (2018) reported that trust was the strongest determinant of attitude, followed by perceived ease of use and perceived usefulness. As for intention, the authors found that attitude showed the strongest influence, followed by perceived usefulness and trust. Likewise, Casal o et al. (2011) found that attitude was the strongest determinant of intention to follow online advice.
The direct positive effect of trust on intention agrees with Anaya-S anchez et al. (2019) and Jalilvand et al. (2017), who showed that trust predicted restaurant online and offline recommendation intentions. However, this result disagrees with Ayeh et al. (2016), who did not support the direct effect of trust on travelers' intention. In conclusion, while previous restaurant research on online reviews adoption labeled attitude (Popy and Bappy, 2020;Salehi-Esfahani and Kang, 2019) or perceived usefulness (Liu et al., 2017) as the strongest determinant of intentions, we extend the literature by revealing that trust is the strongest predictor of the intention to follow online restaurant community advice.

Theoretical implications
Several contributions to theory stand out from our study. First, it answers the calls of DiPietro (2017) and Moreno and Tejada (2019), who revealed that little is known concerning UGC in the restaurant industry compared with hotels, despite its significant role in assisting consumers in making decisions based on online community members' shared experiences. Hence, this work advances the literature by applying the TAM to the context of online restaurant communities to identify the key precursors of the intention to follow online advice.
Second, we contribute to the emerging body of the literature as most studies have focused on the impact of online restaurant reviews on consumer decisions from the perspective of review characteristics or the features of reviewers and readers (e.g. Bigne et al., 2020;Chen and Farn, 2020;Hern andez-Ortega, 2020;Huang and Liang, 2021;Ruiz-Mafe et al., 2020;Zhang et al., 2021). In addition, researchers have examined what motivates customers to write online restaurant reviews and what drive customers' eWOM intentions (e.g. Dixit et al., 2019;Jalilvand et al., 2017;Yang, 2017). This study, however, took a different approach and examined what induces review readers to follow an online advice.
Third, to the best of our knowledge, this research is one of the few to integrate trust into the TAM to explain the extent to which consumers can trust online restaurant community advice and act upon it. The findings contribute to the literature by identifying trust as the most significant driver of consumers' intention to follow online advice, capturing more variance in intention than the technology acceptance factors.
Finally, this study focused on third-party online communities on Facebook for restaurant reviews. Researchers' attention has been mostly drawn to examining online reviews from dedicated online opinion platforms (e.g. TripAdvisor and Yelp) to investigate their influence on consumers' decisions (e.g. Huang and Liang, 2021;Ruiz-Mafe et al., 2020;Srivastava and Kalro, 2019). However, little is known about online reviews written on SNSs such as Facebook (Ladhari and Michaud, 2015).

Managerial implications
Although the focus of the study was on third-party online communities, our findings present practical implications for both restaurant managers and online community administrators.
A predominant finding of this study is that consumers trust online reviews on third-party online communities, and it is the most relevant determinant of their intention to follow the advice. Accordingly, restaurants should encourage their customers to write reviews not only on their own online platforms but also on third-party online communities. Although dissatisfied customers do not hesitate to write online reviews to voice their disappointment, satisfied customers are often unwilling to share their experiences (Huang and Liang, 2021). Therefore, restaurants do not only have to continuously deliver positive dining experiences but should also ensure that these experiences are translated to positive reviews. This can be achieved by offering customers rewards such as discounts or loyalty points to motivate them to share their experiences.
Simultaneously, restaurant marketers should closely monitor consumer reviews on these platforms. They should pay attention to consumers' online feedback and respond to reviews as research reveals that 90% of global diners view online reviews as important when choosing a restaurant, with 63% revealing they are more likely to visit restaurants that respond to reviews (TripAdvisor, 2020). Online reviews reveal a great deal of information about consumers' needs and wants and allow restaurants to provide a more personalized dining experience. Furthermore, monitoring online opinions could reveal weak areas in customers' dining experiences that require improvements and inform restaurants of how they fare against competition.
Consumers have the power to influence peer decisions through eWOM as conventional marketing methods are becoming less influential. Marketers should take advantage of consumers' role as co-promoters of their services and integrate this free advertising in their marketing and communication strategies. Thus, we recommend that restaurants should identify key opinion leaders who frequently contribute to third-party platforms. Opinion leaders exert considerable influence on the decisions of their audience especially when it comes to experience-based services (Ladhari and Michaud, 2015). Once identified, they should be invited to new restaurant openings or when introducing new menus and be encouraged to write honest reviews about their experiences. When these reviews are shared, restaurants would gain exposure on a wide scale because of the popularity of these third-party online communities. In addition, restaurants can display reviews posted by opinion leaders and other satisfied guests on their own websites and social media channels or any offline promotional activities to encourage potential customers to visit them.
As for online community administrators, our results suggest that they should create a trustworthy environment and foster online trust among members. For example, reviews sent by users should be carefully evaluated to ensure the quality and honesty of information posted on the platform. Furthermore, managers can enhance members' trust in the online community by showing them the feedback of satisfied users who adopted the advice in their dining decisions. This could lead to more positive attitudes and would significantly influence their intention to follow the advice.
Finally, taking into consideration the findings of this study, online community designers should focus on creating a user-friendly interface to enhance usefulness perceptions. One way is to make the reviews more searchable. For instance, reviews can be segmented based on type of cuisine or restaurant location by adding hashtags to enable members to easily retrieve reviews that fit their preferences. Another way is to set criteria for writing useful reviews that members can adhere to. Particularly, they can advise members to share more detailed information about different aspects of their experiences including food quality, menu variety, service, atmosphere and location. Moreover, they can encourage reviewers to provide both positive and negative aspects of their dining experience as two-sided reviews are more helpful (Filieri et al., 2018).

Limitations and future research
This work is subject to limitations that serve as directions for future research. First, this work investigates third-party online communities that share consumer-generated restaurant reviews. Future research can test the validity of our model in restaurants' own online platforms and compare the results. Second, the study is conducted in Egypt, which limits the generalizability of our results. Future research can use a sample of respondents with different cultural backgrounds to examine the influence of cultural differences on drivers of the intention to follow online advice. Finally, future research can consider the inclusion of additional constructs to contribute to the unexplained variance. It would be interesting to propose external factors to explain perceived usefulness and perceived ease of use or to propose antecedents for trust in online community.